This paper aims to clarify the usefulness and problems when the neural network applied to epidemiological analysis. The occurrence of hypertension in 10 years was predicted by the neural network and ordinary logistic regression. The neural network model showed sex, age, blood pressure and uric acid as the significant variables to determine the occurrence. On the other hand, the logistic regression showed sex, age, total cholesterol, triglyceride and fasten plasma glucose as the significant variables. The results of the neural network model indicated the data set of the learning samples should be modified to improve learning ability of the neural network.
